fixed ctrl+C on a concurrence container

[athackst]

This change is

[jbohren]

As this PR stands, it has one new busy loop that will easily soak up 100% cpu while the concurrence is executing, and another one that will do the same while it's waiting for the others to terminate. I agree that the issue is with this line but there's nothing significant about replacing the Condition with an Event. You could also just replicate something like this loop on that line, but that would also be sub-optimal

The right way to fix this is to have every Concurrence register its _done_cond or _done_event with whatever function is handling the SIGINT. That way there aren't even idle loops. Alternatively, you can make sure each state is keeping track of the global smach.is_shutdown() and break if that gets set.

[athackst]

So what function is handling the SIGINT for a concurrence?

[jbohren]

In rospy, a SIGINT handler gets registered. What smach should do is provide a global function like smach.register_shutdown_handler(), which would add a given callback to a list of handlers. It shold also provide a global function like smach.handle_shutdown() which would call all of the registered handlers. Then, what smach_ros should do on initialization is call rospy.on_shutdown(smach.handle_shutdown).

[jbohren]

For non-ros use, you can register your own SIGINT handler and set it to call smach.handle_shutdown().

[jbohren]

(: This would be a nice improvement and could get rid of any other loops that exist simply to check smach.is_shutdown()

[athackst]

I agree that sounds better. However I can't seem to get it to work leaving in the that line. It seems like it's blocking the main thread? Nor does simply checking for smach.is_shutdown() in the state seem to break execution with it in.

[athackst]

It was blocking the main thread, which ros uses to update signals like sigint and callbacks. I added a starting function which will spin up a state in another thread and handlers for shutdown. To start so it doesn't block you can just do smach_ros.start(sm), where sm is the state you want tot start.

[athackst]

I also removed the smach.is_shutdown loop in concurrency since it didn't seem necessary any more. All states should now be preempting themselves by default on shutdown and therefore terminate, but I can add it back in if you think it breaks old code.

[athackst]

We decided not to go with smach for our use case, so I'm closing this.

[jbohren]

I'd still like to merge this in with some testing on my end. So I'm going to leave it open, feel free to unsubscribe if you want, though.

[130s]

Closing in favor of #49